Lattice Project Builder turns one product brief into your PROJECT_GUIDE.md, version-locked stack, AI guardrails, CI/CD workflows, and a custom setup command so Cursor, Claude Code, or Codex can build against the same plan.
Turn one product brief into a project plan instead of re-explaining your app in every new AI chat.
Cursor, Claude Code, and Codex get project-specific rules so they follow your architecture and stack.
Get the stack, CI/CD, verification, and the setup command in one package instead of a weekend of setup work.
Project Builder does not stop at a plan. It creates the files each AI editor actually reads, so your prompts, project structure, and code decisions stay aligned even if you switch tools mid-build.
Cursor gets both markdown and structured rules, so it stops improvising versions, folders, and architecture decisions.
Claude Code gets project instructions plus permission and hook scaffolding that match the plan you approved.
Codex gets a clear entrypoint and workflow rules, so it can work inside the same blueprint instead of starting cold.
From idea to a stack your AI can actually follow in under 5 minutes.
Fill out one form: what you're building, project type (web/mobile/API), key decisions (auth, database, UI style), and planning mode.
Takes 2-3 minutes. The form guides you through the important choices.
Lattice AI creates your PROJECT_GUIDE.md and sets up Lattice Core: stack, AI rules, CI/CD workflows, verification, and editor-specific guardrails.
Multi-model deliberation pipeline: draft, critique, and synthesis stages produce a production-ready plan.
Project Builder outputs a single npx command customized for your exact project configuration.
Copy this command. It contains everything needed to set up your project.
Paste the command in your terminal. Everything sets up automatically, and you're ready to code.
Open the repo in Cursor, Claude Code, or Codex and your assistant already has project-specific guardrails.
One brief becomes the plan, the guardrails, and the bootstrapped setup your AI can use immediately.
A shared blueprint you and your AI can return to instead of starting every new chat from scratch.
GitHub Actions run verify on every push so AI-generated drift gets caught before it lands on main.
One paste creates the repo shape, rules, and config for the exact stack you chose.
Dependencies and tooling are chosen as a compatible set, so the AI stops inventing packages or mismatched versions.
Cursor, Claude Code, and Codex each get the files they actually read: .cursor rules, CLAUDE.md, AGENTS.md, and docs/rules/codex.md.
Fast checks keep your human changes and AI changes honest from day one, locally and in CI.
Purple = Project Builder-only features |Green = Includes Lattice Core
One command generates this entire project structure. Every file, folder, and configuration, ready to code.
my-saas/ |-- .cursor/rules.md # AI guardrails for Cursor |-- .cursor/rules/ | |-- lattice-core.mdc # Structured Cursor rules |-- .claude/ | |-- settings.json # Claude Code permissions + hooks | |-- hooks/ # Hook templates for policy checks |-- .mcp.json # MCP server template config |-- .github/ | |-- workflows/ci.yml # CI/CD pipeline | |-- PULL_REQUEST_TEMPLATE.md # PR template |-- .gitmessage # Commit message template |-- .lattice/ | |-- config.json # Lattice config | |-- prompts/ # 6 prompt templates | | |-- 01-feature.md | | |-- 02-bugfix.md | | |-- 03-refactor.md | | |-- 04-test.md | | |-- 05-polish.md | | |-- 06-security.md |-- AGENTS.md # Codex entrypoint + shared execution policy |-- docs/ | |-- rules/ | | |-- codex.md # Codex-specific workflow rules |-- CLAUDE.md # AI guardrails for Claude Code |-- PROJECT_GUIDE.md # Architecture and milestones | |-- app/ | |-- (auth)/ # PHASE 1: WORKING AUTH | | |-- layout.tsx | | |-- login/page.tsx # Login with email/password | | |-- signup/page.tsx # Signup with email confirmation | | |-- forgot-password/page.tsx # Password reset | |-- (dashboard)/ # Protected routes | | |-- layout.tsx # Dashboard layout with nav | | |-- dashboard/page.tsx # Dashboard home | |-- api/ # PHASE 4: API ROUTES | | |-- health/route.ts # Health check | | |-- user/route.ts # User profile CRUD | | |-- example/ | | | |-- route.ts # List/Create | | | |-- [id]/route.ts # Get/Update/Delete | |-- auth/ | | |-- callback/route.ts # OAuth callback | | |-- signout/route.ts # Sign out | |-- layout.tsx # Root layout | |-- page.tsx # Landing page | |-- globals.css # CSS with theme variables | |-- components/ | |-- ui/ # PHASE 3: UI COMPONENTS | | |-- button.tsx | | |-- input.tsx | | |-- card.tsx | | |-- spinner.tsx | | |-- index.ts | |-- lib/ | |-- supabase/ # PHASE 1: AUTH CLIENTS | | |-- client.ts # Browser client | | |-- server.ts # Server client | |-- utils.ts # cn() utility | |-- supabase/ | |-- migrations/ # PHASE 2: DATABASE | | |-- 001_initial_schema.sql # Profiles table + RLS | |-- types/ | |-- database.types.ts # Database types | |-- proxy.ts # Route protection |-- package.json # Full dependencies |-- tsconfig.json # TypeScript config |-- next.config.ts # Next.js config |-- eslint.config.mjs # ESLint flat config |-- jest.config.js # Jest config |-- jest.setup.js # Testing setup |-- next-env.d.ts # Next.js types |-- .env.example # Environment template |-- .env.local # Local env (placeholder) |-- .gitignore # Git ignore |-- README.md # Full documentation
You've felt it. The AI writes fast, but you spend days fixing what it built wrong.
Without clear specs, AI makes assumptions. Wrong assumptions compound into wasted days.
You build, realize it's wrong, tear it down, rebuild. The loop never ends.
AI suggests packages that don't exist, versions that conflict, patterns that are outdated.
Features get bolted on. The codebase becomes unmaintainable. You consider starting over.
Choose the depth that matches your project complexity.
Fast single-pass generation for straightforward projects. Get your custom command quickly.
Multi-model deliberation for complex projects. Multiple AI models debate and refine your setup.
If you want to add guardrails to an existing project, Lattice Core is what you need. One command adds version-aware AI rules, verification scripts, and more. Free forever.
Get Lattice CoreBasic Planning uses a fast, single-pass AI engine, perfect for most projects. Pro Planning uses multiple AI models that debate and refine your setup, catching edge cases and producing deeper architecture analysis. Both are included at every paid tier.
Basic Planning takes 15-30 seconds. Pro Planning takes 60-90 seconds due to multi-model deliberation. After generation, you get your custom command instantly and can start coding in under 5 minutes total.
Each project gives you PROJECT_GUIDE.md, a compatible stack, CI/CD workflows, a custom setup command, verification scripts, and editor-specific rules for Cursor, Claude Code, and Codex. It is meant to be the complete starting point for a new build.
Yes. Project Builder generates the files each tool actually reads: .cursor/rules.md and lattice-core.mdc for Cursor, CLAUDE.md plus .claude/settings.json for Claude Code, and AGENTS.md plus docs/rules/codex.md for Codex.
Project Builder is designed for new projects. If you have an existing project, use Lattice Core (free) to add guardrails. Project Builder generates the complete project structure from scratch.
You can upgrade your plan at any time to get more projects. Unused projects don't roll over, but you can always upgrade mid-month to access higher quotas immediately.
Start with one-time PAYG builds (Basic or Pro), or choose a subscription if you plan to generate regularly.
Turn one product brief into a build-ready repo for Cursor, Claude Code, or Codex and skip the 4-6 hour setup spiral.
PAYG starts at $8 per build. Subscriptions start at $20/month.